This article appeared in a journal published by Elsevier. The attached
copy is furnished to the author for internal non-commercial research
and education use, including for instruction at the authors institution
and sharing with colleagues.
Other uses, including reproduction and distribution, or selling or
licensing copies, or posting to personal, institutional or third party
websites are prohibited.
In most cases authors are permitted to post their version of the
article (e.g. in Word or Tex form) to their personal website or
institutional repository. Authors requiring further information
regarding Elsevier’s archiving and manuscript policies are
encouraged to visit:
http://www.elsevier.com/copyright
Author's personal copy
Available online at www.sciencedirect.com
Advances in Space Research 50 (2012) 1007–1029
www.elsevier.com/locate/asr
A review of global satellite-derived snow products
Allan Frei a,⇑, Marco Tedesco b, Shihyan Lee c, James Foster d, Dorothy K. Hall d,
Richard Kelly e, David A. Robinson f
a
Department of Geography, Hunter College Rm. 1006N, City University of New York, 695 Park Avenue, New York, NY 10065, USA
b
Department of Earth and Atmospheric Sciences, City College, City University of New York, NY 10031, USA
c
Sigma Space Corporation, 4600 Forbes Blvd., Lanham, MD 20706, USA
d
Hydrospheric and Biospheric Sciences Laboratory (Code 617.0), NASA/Goddard Space Flight Center, Greenbelt, MD 20771, USA
e
Department of Geography & Environmental Management, University of Waterloo, 200 University Avenue, West Waterloo, Ontario, Canada N2L 3G1
f
Department of Geography, Rutgers University, 54 Joyce Kilmer Avenue, Piscataway, NJ 08854-8054, USA
Available online 29 December 2011
Abstract
Snow cover over the Northern Hemisphere plays a crucial role in the Earth’s hydrology and surface energy balance, and modulates
feedbacks that control variations of global climate. While many of these variations are associated with exchanges of energy and mass
between the land surface and the atmosphere, other expected changes are likely to propagate downstream and affect oceanic processes
in coastal zones. For example, a large component of the freshwater flux into the Arctic Ocean comes from snow melt. The timing and
magnitude of this flux affects biological and thermodynamic processes in the Arctic Ocean, and potentially across the globe through their
impact on North Atlantic Deep Water formation.
Several recent global remotely sensed products provide information at unprecedented temporal, spatial, and spectral resolutions. In
this article we review the theoretical underpinnings and characteristics of three key products. We also demonstrate the seasonal and spatial patterns of agreement and disagreement amongst them, and discuss current and future directions in their application and development. Though there is general agreement amongst these products, there can be disagreement over certain geographic regions and under
conditions of ephemeral, patchy and melting snow.
Ó 2011 COSPAR. Published by Elsevier Ltd. All rights reserved.
Keywords: Snow; Remote sensing
1. Introduction
Snow covers a considerable portion of Northern Hemisphere lands during winter. It is the component of the cryosphere with the largest seasonal variation in spatial extent.
In fact accumulation and rapid melt are two of the most
dramatic seasonal environmental changes of any kind on
the Earth’s surface (Gutzler and Rosen, 1992; Robinson
and Frei, 2000; Robinson et al., 1993). In the Southern
Hemisphere, outside of Antarctica and its surrounding
ice shelves and sea ice, snow is generally limited to smaller
⇑ Corresponding author. Tel.: +1 212 772 5322; fax: +1 212 772 5268.
E-mail address: [email protected] (A. Frei).
regions such as the Andes, Patagonia and the southern
Alps of New Zealand (Foster et al., 2008). On decadal time
scales, snow variations over Northern Hemisphere lands
have also been considerable (Barry et al., 1995; Brown,
2000; Brown and Braaten, 1998; Derksen et al., 2004; Frei
et al., 1999; Mote, 2006; Mote et al., 2005; Ye et al., 1998),
with declines in spring associated with warmer conditions
(Brown et al., 2010; Groisman et al., 1994; IPCC, 2007;
Leathers and Robinson, 1993). Recent reports on changes
in the Arctic environment cite snow as one of the critical
variables (ACIA, 2004; AMAP, 2011). The expectation
during the 21st century is that changes will be increasingly
dramatic (Frei and Gong, 2005; Raisanen, 2007; Ye and
Mather, 1997) and spatially and temporally complex
(Brown and Mote, 2009; Nolin and Daly, 2006).
0273-1177/$36.00 Ó 2011 COSPAR. Published by Elsevier Ltd. All rights reserved.
doi:10.1016/j.asr.2011.12.021
Author's personal copy
1008
A. Frei et al. / Advances in Space Research 50 (2012) 1007–1029
While large scale changes in snow cover are useful as
indicators of climatic variations, snow also affects other
components of the Earth system at a variety of scales. By
virtue of its radiative and thermal properties which modulate transfers of energy and mass at the surface-atmosphere
interface (Zhang, 2005), snow affects the overlying atmosphere (Barry, 2002; Barry et al., 2007; Cohen, 1994; Ellis
and Leathers, 1999; Mote, 2008; Walsh, 1984) and thereby
plays an important role in the complex web of feedbacks
that control local to global climate. For example, because
of the high albedo of snow, its presence can change the surface energy balance over land and ice and therefore affect
climate (i.e. the snow-albedo feedback). Snow also modulates the hydrologic cycle (Dyer, 2008; Graybeal and
Leathers, 2006; Leathers et al., 1998; Todhunter, 2001);
influences ecosystem functioning (Jones et al., 2001); and
is a significant resource for many mid latitude populations
and for populations whose water is derived from mountainous and northerly cold regions (Barnett et al., 2005;
Barry et al., 2007). Snow observations are critical for the
validation of climate models (Foster et al., 1996; Frei
et al., 2003, 2005; MacKay et al., 2006; Roesch et al., 1999).
With regards to the freshwater flux to the ocean, the role
of snow is to modulate seasonal timing, and in some cases
the amount, of discharge into the oceans. While this can
affect coastal systems across mid-latitudes, of particular
relevance is the fresh water flux into the Arctic basin.
The drainage area into the Arctic Ocean is 1.5 times the
surface area of the Arctic Ocean itself (Peterson et al.,
2002) and river runoff is the largest source of freshwater
input into the Arctic basin (Arnell, 2005; Miller and Russell, 2000). Much of Arctic precipitation is derived from
snow fall, and much of the river runoff is derived from
snow melt. During the past century, both high latitude precipitation (Zhang et al., 2007) and river runoff to the Arctic
basin have increased; both are expected to increase further
in a warming climate (Peterson et al., 2002), although the
rates of change and relative impacts on ocean circulation
vary spatially (Rennermalm et al., 2007).
The studies described above do not include all the possible nonlinear feedbacks in which snow plays a role in
the Arctic environment (Hinzman et al., 2005). For example, due to the insulating effect of snow cover, changes in
the timing of snow onset or disappearance, or changes in
the amount of snow, may influence the state of the underlying permafrost, which has been warming for decades
(Romanovsky et al., 2010) and which is expected to deteriorate during this century (Lawrence and Slater, 2005) and
may further increase the freshwater flux. Thawing permafrost may also result in a significant release of carbon to
the atmosphere as the result of microbial decomposition
of currently frozen organic carbon (Schuur et al., 2008).
According to Betts (2000) the expected expansion of the
boreal forest may lead to both negative feedbacks (an additional carbon sink) and positive feedbacks (an albedo
decrease) on global climate, and the net effect will be a positive feedback with increased warming. The feedbacks
between snow, permafrost, and freshwater flux to the Arctic Ocean associated with these processes are poorly understood (Francis et al., 2009; Rawlins et al., 2010).
While an increased freshwater flux to the Arctic has
potential effects on thermodynamic and ecological processes in the coastal zone, perhaps most importantly such
increases have been shown in the past to diminish or completely halt the formation of North Atlantic Deep Water
(NADW) (Rahmstorf, 2000). This occurs because freshwater export to the North Atlantic Ocean, the region of
NADW formation, decreases surface water density. Model
simulations suggest that the magnitude of expected runoff
changes during this century may approach critical thresholds for NADW formation (Arnell, 2005; Miller and Russell, 2000; Peterson et al., 2002). In a recent study, NADW
formation as well as permafrost degradation and changes
to the tundra and boreal forest ecosystems (all of which
can be affected by snow, and all of which can affect the
freshwater flux to the ocean) have been listed among the
potentially critical components of the Earth system that
may be in danger of approaching “tipping points” (Lenton
et al., 2008). Thus, accurate monitoring of high latitude
snow remains an essential goal.
Because of the large extent of terrestrial snow cover and
the difficulties in obtaining ground measurements over cold
regions, remote sensing represents an important tool for
studying snow properties at regional to global scales. In
recent years, advances in satellite capabilities, as well as
in algorithm development, have led to improved monitoring of snow across the globe. The purpose of this article
is to review the current generation of satellite-derived global snow observations that has become available during
the first decade of the twenty first century, with emphasis
on land surfaces of the Northern Hemisphere. Theoretical
considerations for the remote sensing of snow, and key
products are discussed.
2. Theoretical background
Due to the nature of interactions between snow cover
and electromagnetic radiation of different frequencies,
snow can be distinguished from other terrestrial surfaces
using satellite observations based on a number of different
active and passive techniques (Dozier, 1989; Nolin, 2010).
The two types of instruments used for monitoring global
scale snow variations rely on either (1) a combination of
the visible and infrared, or (2) microwave, portions of the
electromagnetic spectrum (Hall et al., 2005; Matzler,
1994; Rango et al., 2000; Scherer et al., 2005; Schmugge
et al., 2002). These methods are limited by a number of factors, such as clouds, forest cover fraction, terrain heterogeneity and precipitation. For example, interpretation of
visible and infrared as well as passive microwave images
can be difficult where complex terrain causes considerable
spatial variation within each remotely-sensed footprint of
snow depth, surface characteristics, and satellite viewing
angles. Nevertheless, products based on these observations
Author's personal copy
A. Frei et al. / Advances in Space Research 50 (2012) 1007–1029
have been vital for monitoring snow and for our understanding of the role of snow in the Earth system. Though
global active microwave data (e.g., QuikSCAT) can also
be used to study snow extent and depth at relatively large
spatial scale (Tedesco and Miller, 2007a,b), data are available only from 1999 to 2009 (when the satellite failed well
past its expected lifetime; see http://www.jpl.nasa.gov/
news/news.cfm?release=2009-175 downloaded November
2011). In contrast, passive microwave data have been available since the late 1970s, and continue to be available. At
regional scales, airborne data can also be collected before
and after the snow falls to study the attenuation introduced
by the snow pack on naturally emitted gamma radiation
(Carroll, 1987). However, the data collected with this
method have low temporal resolution (seasonal scale) and
cannot be used for global scale studies. Consequently, we
focus our analysis on snow parameters estimated by means
of visible and infrared and passive microwave sensors.
2.1. Visible and near-infrared
Snow extent (i.e. presence or absence of snow, regardless
of snow amount) is, in many circumstances, relatively
straightforward to observe using visible observations
because of the high albedo of snow (up to 80% or more
in the visible part of the electromagnetic spectrum) relative
to most land surfaces. However, limitations exist. First, visible imagery is limited to that portion of the surface illuminated by sunlight; thus darkness and low illumination are
problematic. Second, clouds impede visible evaluation in
two ways. All but the thinnest clouds reflect a significant
portion of visible radiation, preventing any visible radiative
information about the surface from reaching the satellite.
And, because the albedos of clouds and snow are often similar, the discrimination between cloud-covered and snow
covered surfaces can be difficult. However, near-infrared
bands can be used to distinguish between snow and most
clouds because the near-infrared reflectance of most clouds
is high while the near-infrared reflectance of snow is low.
Third, vegetation can obstruct visible and infrared information about snow from reaching the satellite sensor. Forest canopies protrude above the snow pack, lowering the
surface albedo (Robinson and Kukla, 1985) and partially
or completely obscuring the underlying surface, making it
difficult to determine snow extent or amount (Chang
et al., 1996; Derksen, 2008; Goita et al., 2003; Klein
et al., 1998; Nolin, 2004).
Lastly, surface heterogeneity can play a role the interpretation of visible and infrared imagery in a number of
ways. Of particular relevance to the monitoring of high latitude snow is the presence of numerous frozen lakes in Arctic regions, which may contribute to the overestimation of
snow covered area from visible and infrared based imagery
during periods when lakes remain frozen after the snow has
melted on adjacent land surfaces (Derksen et al., 2005a;
Frei and Lee, 2010), at least when high resolution land surface data sets are not used to filter out the signal from lake
1009
surfaces. Passive microwave based estimates of SWE may
be underestimated due to the presence of lakes (Derksen
et al., 2005a; Rees et al., 2006). On the other hand, surface
heterogeneity may assist in the interpretation of snow-covered versus snow-free ground, and of snow-covered versus
cloud-covered scenes, when trained analysts are mapping
snow extent using visible imagery.
2.2. Passive microwave
Because snow grain dimensions can be similar to microwave wavelengths, snow is efficient at scattering the microwave radiation naturally emitted from the Earth’s surface
(Matzler, 1994). Therefore, microwave emission from a
snow covered surface is diminished relative to a snow-free
surface, and the presence of snow can frequently be identified (Chang et al., 1976; Grody, 2008; Hall et al., 2005;
Matzler, 1994; Tait, 1998; Tedesco and Kim, 2006). Furthermore, because under ideal circumstances the amount
of scattering is proportional to the number of snow grains,
microwave instruments offer the possibility of estimating
the mass per unit area of water in the snow pack, which
is often measured as snow water equivalent (SWE). In contrast to visible and infrared, passive microwave does not
depend on the presence of sunlight and thus provides an
alternative at high latitudes; and, passive microwave is largely (but not completely) transmitted through non-precipitating clouds, offering the potential to estimate snow cover
under many cloudy conditions that preclude visible and
infrared observations. In practice, research using passive
microwave exploits the fact that microwave scattering by
ice crystals is frequency-dependent: higher frequencies
within the microwave portion of the spectrum are scattered
more efficiently than lower frequencies, enabling the use of
two or more frequency bands to estimate SWE (Chang
et al., 1987; Derksen, 2008; Derksen et al., 2005b; Grody
and Basist, 1996). Other methods have also been evaluated
such as one based on the inversion of a snow emission
model (e.g., Pulliainen and Hallikainen, 2001). Clifford
(2010) provides a review of global estimates of snow water
equivalent from passive microwave.
Limitations to the monitoring of snow using passive
microwave sensors are due to a variety of factors. One
major limitation is the presence of liquid water in the snow
pack, the microwave emission from which masks the snow
signal and inhibits the ability of microwave sensors to
detect wet snow. Also, because of the relatively weak
microwave signal emitted by terrestrial surfaces, microwave sensor footprints are necessarily large (25 km).
Uncertainties in snow depth and SWE estimates are associated with the physical structure of snow packs (ice lenses,
grain size variations and vertical heterogeneity) which vary
in space (Chang et al., 1976; Sturm et al., 1995) and time
(Langham, 1981) and can alter the scattering and emission
characteristics of the snow pack. Snow pack metamorphosis, which in the Arctic region typically results in a layer of
depth hoar (with large crystal size) near the bottom of the
Author's personal copy
1010
A. Frei et al. / Advances in Space Research 50 (2012) 1007–1029
pack, results in more efficient microwave scattering. Thus,
a signal change measured at the satellite sensor due to snow
metamorphosis can mimic a signal change due to a change
in SWE. Vegetation in and above the snow pack emits
microwave radiation, and can confound any detection
algorithm (Chang et al., 1996; Foster et al., 1997; Tedesco
et al., 2005).
Finally, as a snow pack reaches a certain critical depth
the relationship between snow-amount and MW brightness
temperature reverses (Derksen, 2008; Markus et al., 2006;
Matzler, 1994; Schanda et al., 1983). When SWE exceeds
150 mm emission by the snow pack of microwave band
radiation is greater than scattering, resulting in a positive
relationship between SWE and brightness temperature.
This is an additional source of uncertainty in SWE retrievals for deep snow packs.
2.3. Remote sensing of snow in complex terrain
Some of the difficulties inherent in the interpretation of
remotely sensed images are exacerbated in regions with
complex terrain (Dozier, 1989). Due to variability of slope,
aspect, and land cover, the local solar illumination angle
varies within one satellite footprint. In fact, due to co-registration differences between an image and a digital elevation model, illumination angles, and therefore reflectance
characteristics, are often unknown. To address such issues,
Painter et al. (2009) developed the MODSCAG model,
which estimates mean grain size and fractional snow cover
from MODIS data using linear spectral mixture analysis
and a library of reflectance characteristics of different surface types. This model has relatively small errors, and
could potentially be applied globally, but so far has been
validated mostly in regions of complex terrain.
A recent study of different algorithms for estimating
SWE from passive microwave radiances in a basin with
complex terrain in the Canadian Rockies finds that the traditional algorithms which are based on brightness temperature differences in difference wavelength intervals (as
discussed above) are less accurate than Artificial Neural
Network (ANN) techniques which can be trained on observations from surface stations (Tong et al., 2010a). Unfortunately, due to the limited distribution of stations for
training the ANN in their test region, they are unable to
accurately estimate spatial variations of SWE. Tong and
Velicogna (2010) and Tong et al. (2010b), using surface station observations and MODIS imagery across the Mackenzie River Basin, determine that the minimum or threshold
SWE value estimated from passive microwave observations
that can be used to determine the presence/absence of snow
varies from sub-basin to sub-basin, depending on topography and vegetation cover. Nevertheless they find useful
information in remotely sensed SWE values for hydrologic
monitoring. As these studies indicate, the estimation of
snow characteristics in complex terrain from remotely
sensed imagery is an important and cutting edge field of
study. At this time, these techniques have not been incor-
porated into global products, and are not addressed further
in this paper.
2.4. Comparison and evaluation of products
When two products disagree, which is “correct?” Is
either one of them “correct?” Two key impediments to a
conclusive evaluation are that there is no perfect “ground
truth,” and that the answer depends on spatial scale. The
most obvious method of testing the veracity of remotely
sensed (or other gridded) products is by comparison to surface reference observations. However, there exists considerable contrast between surface, or in situ, and remote snow
observations with regards to the snow pack properties that
can be measured, their spatial and temporal resolutions
and domains, and the methods employed to make measurements (Brown and Armstrong, 2008; Goodison et al.,
1981).
Even in regions with surface observations, validation
may be difficult because of the contrasting spatial scales
of surface and remotely sensed observations (Brubaker
et al., 2005; Chang et al., 2005). Brubaker et al. (2005) discuss the difficulties in comparing point measurements to
spatially integrated satellite retrievals, especially in areas
of sparse station networks, which are typically at high elevations and northerly regions (exactly the areas where
snow is most prevalent). They find that there is no single
accepted method to perform validation of remotely sensed
snow products. Chang et al. (2005) provide an informative
review of how varying station densities and different satellite footprints are not equally spatially representative, and
how the differences can complicate evaluations and comparisons of different products. They employ geostatistical
techniques, as suggested by Kelly et al. (2004), to quantitatively define the spatial density of station observations
required to provide sufficient information for validation
studies. MODIS has been found to compare well with station based observations as well as with the National Operational Hydrologic Remote Sensing Center products (Hall
and Riggs, 2007), but Riggs et al. (2005) show that even
between different versions of MODIS snow products, analyses at different spatial resolutions can provide conflicting
results in some cases, due to both the resolution differences
and the averaging method.
Despite the inherent difficulties, comparative studies to
date have drawn some useful conclusions (Armstrong
and Brodzik, 2001; Basist et al., 1996; Bitner et al., 2002;
Brown et al., 2007, 2010; Derksen et al., 2003; Drusch
et al., 2004; Foster et al., 1997; Mialon et al., 2005; Mote
et al., 2003; Romanov et al., 2002; Savoie et al., 2007; Tait
and Armstrong, 1996). For example, evaluations of NOAA
visible and infrared versus passive microwave products find
more disagreement during fall and spring than during midwinter, with particular differences under forest canopies,
over complex terrain, in areas of persistent clouds, patchy
snow, and wet snow (Armstrong and Brodzik, 2001; Basist
et al., 1996). Over the Tibetan Plateau these products often
Author's personal copy
A. Frei et al. / Advances in Space Research 50 (2012) 1007–1029
disagree year-round (Armstrong and Brodzik, 2001; Savoie
et al., 2007). Several recent studies identify differences
between remotely sensed products and surface observations
over North America during the spring ablation season
(Brown et al., 2007, 2010; Frei and Lee, 2010).
3. Snow products
A number of digital products that are based on remote
observations are available. The two visible and infrared
based suites of products that are most widely used for
large-scale climate research are from: (1) the Interactive
Multisensor Snow and Ice Mapping System (IMS) (Section
3.1) and (2) the suite of products derived from the Moderate Resolution Imaging Spectroradiometer (MODIS) (Section 3.2). IMS is the most recent version of a product
that dates back to the 1960s (Matson and Wiesnet, 1981).
IMS mapping of snow extent has relied primarily on visible
and near infrared imagery, but includes data and information from a number of sources. As discussed in more detail
below, the key feature that distinguishes IMS from other
products is human involvement in the analysis, which is
required for operational purposes.
The MODIS instrument, which is used to observe a
number of geophysical variables including snow, flies on
NASA’s Earth Observing System (EOS) Terra satellite,
launched in 1999. A near-twin MODIS instrument is also
flying on board the Aqua platform, which was launched
in 2002 (Aqua and Terra have afternoon and morning
equatorial crossing times, respectively). Aqua also hosted
the Advanced Microwave Scanning Radiometer – Earth
Observing System (AMSR-E) until its failure in October
2011. AMSR-E measured the naturally emitted radiation
in the microwave region at five different frequencies (6.9,
10.7, 18.7, 23.8, 36.5 and 89 GHz) at both vertical and horizontal polarizations.
The IMS and MODIS snow algorithms both rely primarily on near-polar orbiting satellites, from which daily
images are available at high latitudes. Other algorithms
that have been suggested (Romanov et al., 2003; Siljamo
and Hyvarinen, 2011) rely on geostationary satellites,
which have the advantage of higher temporal resolution,
but have poor spatial resolution.
3.1. visible and near infrared based products
3.1.1. The Interactive Multisensor Snow and Ice Mapping
System (IMS)
The data set that has historically been the most widely
used for the operational mapping and climatological analysis of large-scale snow extent (not depth or water equivalent) was produced by the US National Oceanic and
Atmospheric Administration (NOAA) National Environmental Satellite and Data Information Service (NESDIS),
but has been transferred to the National Ice Center
(NIC), which is jointly supported by NOAA, the US Navy,
and the US Coast Guard. This product has been based pri-
1011
marily on visible and near infrared observations, and covers the period from late 1998 to present (Ramsay, 1998),
with the precursor maps beginning in 1966, constituting
the longest remotely sensed environmental time series that
has been derived in a near-consistent fashion (Helfrich
et al., 2007; Matson and Wiesnet, 1981; Robinson et al.,
1993). The term near-consistent is used because, due to
changing operational requirements and evolving technical
capabilities, this product has undergone, and continues to
undergo, improvements and refinements (Helfrich et al.,
2007; Ramsay, 1998, 2000) as summarized briefly here.
The two reasons for this product’s importance – as operational input into atmospheric forecast models and as a
long-term climatic record – are also discussed.
Although a number of improvements and corrections in
the production of the NOAA product occurred in the earlier years, the biggest methodological change was implemented in the late 1990s. Until that time, NOAA snow
maps were produced on a weekly basis by trained meteorologists who would visually interpret photographic copies
of visible band imagery, and manually produce maps that
would subsequently be digitized with spatial resolution
between 150 km and 200 km. In 1997 NOAA began producing snow maps using the IMS, with improved spatial
(24 km) and temporal (daily) resolutions. IMS is operated
by trained analysts who produce a daily digital product utilizing Geographic Information System technology and
incorporating a variety of, and an ongoing expansion of,
technological capabilities as well as sources of information.
Since 1999, when weekly manual mapping was discontinued, daily IMS maps have been produced (Ramsay, 1998;
Robinson et al., 1999). Technological advancements since
1999 have led to even higher resolution (4 km) snow mapping (Helfrich et al., 2007).
IMS produces estimates of snow extent across the globe
every day, regardless of the presence of clouds. This is possible primarily for two reasons. First, analysts use sources
of information other than visible and near infrared imagery. Second, because IMS analysts can loop through
sequential images, their ability to evaluate scenes is based
on an integration of information from both spatial and
temporal perspectives. Thus, a key feature of the IMS
product is that human judgment as to which data sources
are most reliable in different conditions and regions, and
as to the final evaluation of where the snow is, remains
an integral part of the process, and one of the strengths
of the IMS product. IMS also includes sea ice extent, which
is not discussed in this report. Fig. 1 shows an example of a
daily IMS snow map in its original projection.
It is difficult to optimize this product for both of its two
main uses. Its primary purpose is to provide input to atmospheric forecast models. For this purpose, continued
product improvements are advantageous. As a record for
evaluating long term environmental change, however, the
value of any product is diminished if methodological
changes (including those that provide more accurate maps)
result in temporal inconsistencies in the data set that might
Author's personal copy
1012
A. Frei et al. / Advances in Space Research 50 (2012) 1007–1029
Fig. 1. Snow extent on February 10, 2011, obtained from IMS.
be difficult to distinguish from actual variations in snow
extent. To maintain product continuity and a viable longterm record, IMS continues to produce a coarse (24 km)
resolution version of the data set. And, in collaboration
with NOAA, the Rutgers University Global Snow Lab is
producing a climate data record in which inconsistencies
between the earlier maps and the IMS product (in addition
to inconsistencies during the weekly map era) are accounted
for, and can therefore be used for ongoing analyses of historical variations (Robinson and Estilow, 2008).
3.1.2. The Moderate Resolution Imaging Spectroradiometer
(MODIS)
The MODIS sensor measures radiation in 36 spectral
bands, including the visible, near infrared, and infrared
parts of the electromagnetic spectrum. The suite of
MODIS snow cover products, available since 2000, are
derived using a fully-automated algorithm that provides
good spatial resolution (500 m), cloud detection, and frequent coverage (daily at mid to high latitudes) (Hall and
Riggs, 2007; Hall et al., 1995, 2002; Riggs et al., 2006).
The MODIS snow-mapping algorithm uses a normalized
difference between MODIS band 4 (5.45–5.65 mm) and 6
(1.628–1.652 lm) and many additional spectral and threshold tests. In forested areas the threshold is changed based
on results of a canopy reflectance model, using both the
Normalized Difference Vegetation Index (NDVI) and
Normalized Difference Snow Index (NDSI) (Klein et al.,
1998). A thermal mask is also included to remove erroneous “snow” over locations where the presence of snow is
considered to be implausible. See Riggs et al. (2006) and
Riggs and Hall (in press) for a description of the algorithm
and recent upgrades.
NASA provides a hierarchy of snow products based on
MODIS observations, designed to satisfy the needs of a
variety of users (http://modis-snow-ice.gsfc.nasa.gov/).
These include a Level-2 swath product; daily and 8-day
composite Level-3 tile products which are mapped onto
a sinusoidal projection and available in 10° lat/lon tiles;
as well as daily, 8-day composite, and monthly Level-3
products available in the Climate-Modeling Grid (a latitude-longitude grid) at 0.05° resolution (Hall et al., 2002,
2005; Riggs et al., 2005, 2006). An 8-day composite is considered useful because in many regions, particularly at
high latitudes, persistent cloudiness limits the number of
days available for surface observations (see results section). The Climate-Modeling Grid was developed to be
useful for the evaluation of climate models and for studies
at large spatial scales. Other features of the MODIS snow
product suite include daily snow albedo (Klein and Stroeve, 2002) and fractional snow cover (Salomonson and
Appel, 2004). In addition, a new cloud-gap filled product
provides information on cloud persistence, and uses observations from prior days to map snow (Hall et al., 2010).
Fig. 2 shows examples of MODIS snow cover maps in
swath format (MOD10_L2) following a major snowstorm
in the northeastern United States in December 2010. The
analysis presented in Sections 4 And 5 of this article uses
the MOD10C1 daily Level 3 global .05° daily ClimateModeling Grid product with a spatial resolution of
5 km.
Validation of the suite of MODIS snow cover products
has been undertaken by many authors as described in Hall
and Riggs (2007). These products have also been used
extensively in modeling efforts, both at the regional and
hemispheric scales. A bibliography of papers utilizing the
MODIS snow cover products for both validation and modeling may be found at: http://modis-snow-ice.gsfc.nasa.gov/?c=publications.
3.2. Passive microwave based products
Historical passive microwave measurements are available from the Scanning Multichannel Microwave Radiometer (SMMR) instrument (1978–1987), and the Special
Sensor Microwave/Imager (SSM/I) instrument (1987
through present) although some compatibility issues
between the two products exist, due to slight differences in
the frequency bands measured, overpass time, swath width,
native footprint resolution, and coverage issues related to
SMMR being powered down every other day. (Armstrong
and Brodzik, 2001; Brodzik et al., 2007; Derksen and
Walker, 2003). AMSR-E, available from 2002 to October
2011, provides a suite of measurements to make it spectrally
Author's personal copy
A. Frei et al. / Advances in Space Research 50 (2012) 1007–1029
1013
Fig. 2. Images acquired on 28 December 2010 showing extensive snow cover as the result of a major snowstorm in the northeastern United States on 26–
27 December. The image on top is a “true-color” image from the MOD09 reflectance product; the center image is a binary MOD10_L2 swath snow map;
and the image on the bottom represents fractional snow cover from MOD10_L2, so that various shades of blue indicate different fractions of snow cover
within each pixel. Image processing by Jeff Miller, Wylie Systems, and NASA/GSFC.
compatible with both SMMR and SSM/I at higher spatial
resolution (Derksen et al., 2005b; Kelly et al., 2004). Due
to the inherent difficulties and regional variations in the
interpretation of passive microwave signals, the production
of a data set that is consistently accurate across all Northern
Hemisphere regions requires either (1) a physical approach,
which includes robust representations of snow pack pro-
cesses and their parameterization in retrieval schemes
(Pulliainen and Hallikainen, 2001), or (2) a regional
approach, which includes regionally-tuned algorithms (Foster et al., 1997) that statistically represent regional snow
pack processes but which are not applicable in different
snow accumulation regimes. The physical approach is very
challenging yet has the potential of being widely applicable
Author's personal copy
1014
A. Frei et al. / Advances in Space Research 50 (2012) 1007–1029
as our knowledge of, and ability to model, regional snow
pack processes improves, while the statistical approach is
applicable only in the few regions for which retrieval
schemes have been calibrated.
The global AMSR-E SWE product suite (Tedesco et al.,
2011 updated) consists of daily, pentad (five-day) maximum and monthly average SWE estimates that together
comprise the only NASA satellite-based SWE product
available to the scientific community. As an example,
Fig. 3 shows the snow depth obtained from the AMSR-E
product for January 30th, 2005.
The AMSR-E SWE operational algorithm takes advantage of several AMSR-E channels that are unavailable
from SSM/I and SMMR. Snow depth is derived as a combination of microwave brightness temperature differences
at different frequencies, weighted by coefficients derived
from the difference between vertical and horizontal polarizations. These coefficients replaced a previously used static
coefficient to attempt to capture the spatio-temporal variability of parameters such as grain size (Kelly, 2009; Kelly
et al., 2003; Tedesco and Narvekar, 2010). The algorithm
uses a spatially varying but temporally static map of snow
density.
Environment Canada (http://ccin.ca/cms/en/socc/snow/
swe/currentSnow.aspx) also produces a regional passive
microwave SWE product for central Canada, including
the Prairies and part of the boreal forest back to 1978.
Until December 1999, this product relied on a single algo-
Fig. 3. Snow depth on January 30, 2005 obtained from the AMSR-E
product.
rithm that was calibrated for the prairies region (Goodison
and Walker, 1995). Since that time algorithms that correct
for the effects of different forest types (Derksen, 2008;
Goita et al., 2003) and the sub-Arctic tundra (Derksen
et al., 2010) have been developed.
3.3. Combined products
One promising avenue, and an area where great efforts
are currently being made, is to refine our abilities to combine ground observations and models with space borne
remotely sensed data. Tait et al. (2000) provide a helpful
review, and describe a prototype of a fully automated product that includes station observations as well as both visible
and microwave retrievals. Here we briefly review some
products that include combinations of satellite, station
observations, and models. While not exhaustive, it provides
examples of the variety of integrative products and methods that have been produced.
3.3.1. The Canadian Meteorological Center snow product
(station observations and modeling)
Snow depth from the Canadian Meteorological Center
Daily Snow Depth Analysis includes a hybrid modeling/
observational approach based on optimal-interpolation of
daily snow depth observations from hundreds of stations
globally, with snow density estimated from a simple snow
pack model (Brasnett, 1999). This model output is considered most dependable over regions with significant station
coverage, which is generally south of 55° North, where
model results are well constrained by observations. Over
most of the Arctic, in contrast, where there are few observations, the analysis is based mostly on model results, and
is skewed towards snow depth observations at coastal locations with observing sites at open areas near airports. Snow
at these sites tends to be shallower and to melt out earlier
than snow in surrounding terrain. Nevertheless, this analysis is considered to be a reasonable estimate of snow depth
over data-poor Arctic regions, and has been used in a number of studies (Brown and Mote, 2009). Here we use CMC
modeled snow depths for comparison with AMSR-E snow
depths.
3.3.2. GlobSnow (satellite, station, and model)
In 2008 the European Space Agency embarked on an
effort to develop a long term snow cover data set called
GlobSnow with sufficient homogeneity to be acceptable
for climate change analysis. The GlobSnow product currently includes global gridded information on snow extent
and SWE across the Northern Hemisphere (excluding
mountainous regions) (Pulliainen, 2010). The SWE product is based on the method of Pulliainen (2006). By incorporating station observations and snow pack modeling into
passive microwave retrieval algorithms, the goal is to provide an accurate product useful for analyses at many different spatial scales, and for near-real time as well as
climatological studies. The snow extent product is created
Author's personal copy
A. Frei et al. / Advances in Space Research 50 (2012) 1007–1029
using European Space Agency satellite visible and infrared
observations (ERS-2 ATSR-2 and Envisat AATSR) based
on the method of Metsamaki et al. (2005), and will likely be
available at a variety of spatial resolutions. GlobSnow is
currently available (http://www.globsnow.info/) but is
new, so there is little peer-reviewed literature on it at the
time of this writing (Takala et al., 2011).
3.3.3. Other combined products
A variety of combined products have been produced
globally, regionally, or for specific purposes. One widely
used combined product is NOAA’s National Operational
Hydrologic Remote Sensing Center Snow Data Assimilation System, which operationally incorporates input from
snow models, station reports, and airborne sensors to estimate daily SWE at 1 km resolution across the continental
US (Carroll et al., 2001). The product by Brown et al.
(2003), which employs the operational snow depth routine
of the Canadian Meterological Center model (Brasnett,
1999), has been used for evaluation of climate models (Frei
et al., 2005). Foster et al. (2008) recently produced a global
product blending visible and infrared, passive microwave,
and active microwave scatterometer data, with the intention of incorporating the most reliable aspects of each
product. Derksen et al. (2004) produced a product going
back to the early 20th century for the North American
Prairies and Great Plains based on passive microwave
and station observations. Biancamaria et al. (2011) estimated Northern Hemisphere fields of SWE based on passive microwave combined with a dynamic snow grain
model. Grundstein et al. (2002) developed a research-oriented SWE climatology for the Great Plains of the United
States by combining station observations with the 1-dimensional snow pack model SNTHERM (Jordan, 1991). A
research-oriented product based on spatial interpolation
of in situ depth measurements over North America (Dyer
and Mote, 2006) has been used for process studies (Ge
and Gong, 2008). The QuickSCAT active microwave scatterometer has been used to estimate the timing of snow melt
across Greenland (Nghiem et al., 2001) and across Arctic
lands (Wang et al., 2008).
4. Methodology to compare and contrast products
In this section we describe the methodology that we use
to demonstrate the regions over which the products typically differ. This analysis is not meant to provide insight
into new remote sensing techniques, but rather to demonstrate the spatial extents and magnitudes of the differences
between products during different seasons. The methodology employed here is designed to achieve two goals: (1)
to identify regions across the Northern Hemisphere where
there is agreement/disagreement between the three main
products discussed here during clear days; and (2) to provide an indication of the spatial distribution of uncertainty
in the AMSR-E snow depth estimates, as determined by
comparison to the CMC product. In this report we show
1015
results for three months: October (a month of rapid
average increase of snow area), January (the month of largest average snow area), and April (a month of rapid average decrease of snow area). For our analysis, AMSR-E
SWE values are converted to depth. This is done using a
fixed density mask, which is also used as part of the standard product algorithm to estimate SWE values. We
reverse the process in order to convert SWE values to
depth. The reprojection methods, and the methods for each
goal, are discussed in more detail in Sections 4.1–4.3.
Many of our methods for goal 1 closely follow Frei and
Lee (2010), and the reader is referred to that article for
more details and justification of the methods. Note that,
without independent verification, agreement between products does not guarantee that they are correct; and, that if
two of the three products agree, it does not guarantee that
the third product is incorrect.
4.1. Reprojection procedure
IMS and MOD10C1 data sets were reprojected to the
EASE-Grid 25 km projection (Brodzik and Knowles,
2002) (AMSR-E is already in this projection). Each
EASE-Grid cell value was calculated as a binary (i.e. snow
or no-snow) value. Because reprojection can introduce
spurious errors at the grid-cell scale, and these errors
are likely to be exacerbated in areas of variable terrain,
we show no results for EASE-Grid cells within which the
GTOPO 30 DEM elevation field has a standard deviation > 100 m. We also avoid drawing conclusions from
individual grid points, but rather focus on results over large
regions with relatively little topographic variation. The
reprojection, binary snow value calculation, and terrain
masking were performed according to the method of Frei
and Lee (2010).
4.2. Agreement/disagreement between IMS, MODIS, and
AMSR-E snow extent
Since both IMS and MOD10C1 provide binary values
indicating either the presence or absence of snow (the standard MODIS products also provide fractional snow cover)
but not snow depth, AMSR-E snow depths were converted
to a binary value to facilitate this comparison. All AMSRE depth estimates below 5 cm are considered snow free as
that is the depth value assigned to shallow snow.
All snow extent analyses include, at each grid cell, only
days with “clear” skies, and only days for which all three
products have non-missing data. We use the MOD10C1
cloud mask to identify EASE-Grid cells that are mostly
clear. Because MOD10C1 0.05 degree cells are higher resolution than the EASE-Grid and include fractional cloud
cover, they can be used to estimate fractional cloud cover
within each grid cell of our analysis. And, because the
MOD10C1 cloud mask is considered conservative (in the
sense that cloud-covered scenes are unlikely to be designated as “clear”) (Riggs and Hall, 2002), we feel confident
Author's personal copy
1016
A. Frei et al. / Advances in Space Research 50 (2012) 1007–1029
that information from cloudy days is not being retained for
analysis. This is achieved by retaining for analysis, for each
EASE-Grid cell, only days with >80% of MOD10C1 cells
that are <20% cloud covered, for which no product is missing data. Frei and Lee (2010) present the rationale for this
method and explain how the results are insensitive to reasonable values of these parameters. For each grid point
on each day, either all three products agree (i.e. either snow
or no-snow), or one product differs from the other two.
The figures summarizing our results show, for each month,
where each product disagrees with the other two products.
4.3. Comparison of AMSR-E and Canadian Meteorological
Centre (CMC) snow products
For passive microwave data and the CMC model, no
cloud mask is invoked, so we retain for analysis all available
dates. While passive microwave data are not limited by
most clouds, clouds with high liquid water content can
affect the comparison between spaceborne- and groundbased SWE estimates (Wang and Tedesco, 2007); this issue
is ignored here in order to increase the sample size.
The comparison of AMSR-E to the CMC snow product
is done by comparing climatological maps (2003–2010).
For each month, three panels are shown containing maps
of AMSR-E snow depth, CMC snow depth, and the difference between the two products (we calculate the difference
as CMC minus AMSR-E, so that a negative difference indicates that AMSR-E overestimates snow depth with respect
to CMC).
5. Results
In this section we show the results of our analysis, the
purpose of which is to demonstrate the spatial patterns
of disagreement between the data sets. We also discuss possible reasons for disagreements. In some cases these reasons
may be speculative.
5.1. Number of days per month available for analysis
Before discussing disagreements between the products,
we first show maps of the number of days per month available for comparison (Fig. 4) which demonstrate the problem presented by clouds. During October and January
(Fig. 4a and b) most Arctic land surfaces are colored green1
or dark blue, which indicates that on average less than
three (green) or three to six (dark blue) days per month
are available for analysis. (In January one also sees the
“ring” around the Arctic with no data associated with no
solar illumination.) During spring, which tends to be less
cloudy over most regions (Fig. 4c), one can find large por1
For interpretation of color in Fig. 4, the reader is referred to the web
version of this article.
tions of the Arctic with either six to nine or nine to 12 days
per month available for comparison.
The vast majority of the unavailable days are caused by
clouds, not by data that is missing for some other reason.
Any satellite product based on visible and infrared band
radiances will lack information from the surface under
clouds. While passive microwave based products can provide information under most types of clouds, they are
currently unreliable under a number of circumstances (discussed in the next section). Considering the importance of
having daily real-time information about the surface to
specify boundary conditions in weather prediction models,
as well to track climatological changes in snow extent,
IMS, or an equivalent product that provides information
for all days regardless of cloud conditions, is a necessity.
5.2. Disagreement between AMSR-E and the other two
products
Fig. 5 includes, for each month, a map showing where
AMSR-E identifies snow to the exclusion of the both
MOD10C1 and IMS (Fig. 5a, c, e) and a map showing
where AMSR-E finds no snow when the other two products identify snow (Fig. 5b, e, f). The most prominent feature is the red colored plateau region of central Asia seen in
all maps down the left hand column (Fig. 5a, c, e). This
indicates that during all months over this region AMSRE identifies snow more often than the other two products.
While we do not, in general, assume that a product is
wrong because it disagrees with the other two products,
in this region we know from other studies that AMSR-E
observations are biased due to problems in passive microwave snow detection at higher elevations associated with
atmospheric influences on passive microwave retrievals
(Wang and Tedesco, 2007). Since the atmosphere over
the high elevation plateaus is much thinner, the algorithms
calibrated globally at lower elevations require correction
(Savoie et al., 2009).
Panels on the right side of Fig. 5 show that during each
month there are regions where AMSR-E identifies snow
less frequently than MOD10C1 or IMS (Fig. 5b, d, f).
The regions shown on these panels coincide with boundary of the continental snow cover during each month
(see the Rutgers University Global Snow Lab web site
for climatological maps of monthly snow cover based
on IMS: http://climate.rutgers.edu/snowcover/index.php).
Regions near the boundary tend to have patchy, shallow
snow packs. During spring (Fig. 5f) the disagreement
across well defined ablation bands at the southern boundary of the continental snow pack is also likely due to significant areas of melting snow with liquid water in the
snowpack.
5.3. Disagreement between IMS and the other two products
Fig. 6 demonstrates that the most prominent circumstance under which IMS disagrees with the other two
Author's personal copy
A. Frei et al. / Advances in Space Research 50 (2012) 1007–1029
1017
Fig. 4. Average number of days per month at each grid point during which skies are clear and all three products are available (i.e. non-missing data)
during (a) October; (b) January; (c) April.
products is during the spring ablation period near the
boundary of the continental snow pack (Fig. 6e). This
result is in agreement with recent studies (Brown et al.,
2007, 2010; Frei and Lee, 2010) which find that over the
last decade or so the timing of spring ablation over North
America is later, by up to several weeks in the central
Canadian Arctic, according to IMS in comparison to other
observations. The reasons for these discrepancies, which
are found during the entire spring ablation season (April,
May, and June; May and June not shown here) over the
boreal forest as well as the tundra, are not understood,
but may be related to geographic factors such as the forest
type and/or the presence of numerous lakes in this (Derksen et al., 2005a; Rees et al., 2006). Investigations into the
cause of this problem continue.
5.4. Disagreement between MOD10C1 and the other two
products
The most interesting example of MOD10C1 disagreeing
with IMS and AMSR-E is found during autumn over
Eurasia. During October over a broad, seemingly incoherent region of Eurasia, predominantly over Scandinavia and
northern Europe, MOD10C1 often identifies snow when
the other two products do not (Fig. 7a). However, this
region is not as incoherent as it may seem, as it corresponds
closely to the boreal evergreen needleleaf forest as defined
by analysis of MODIS reflectance (Friedl et al., 2002). During November (not shown) we find a similar pattern, except
the differences are more extreme and concentrated more
over Scandinavia. The eastern Eurasian region over which
Author's personal copy
1018
A. Frei et al. / Advances in Space Research 50 (2012) 1007–1029
Fig. 5. Percentage of available days during which AMSR-E disagrees with IMS and MODIS. (a) October; AMSR-E shows snow while other two products
show no snow. (b) October; AMSR-E shows no snow while other two products show snow. (c) January; AMSR-E shows snow while other two products
show no snow. (d) January; AMSR-E shows no snow while other two products show snow. (e) April; AMSR-E shows snow while other two products
show no snow. (f) April; AMSR-E shows no snow while other two products show snow.
MOD10C1 often fails to identify snow when IMS and
AMSR-E see snow (Fig. 7b) corresponds closely to the
region of deciduous needleleaf forest. It seems that over
one type of forest MODIS sees snow more often, while
Author's personal copy
A. Frei et al. / Advances in Space Research 50 (2012) 1007–1029
1019
Fig. 6. Same as Fig. 5 except showing where IMS disagrees with other two products.
over a different type of forest MODIS sees snow less often.
While the difficulties of remotely sensing snow under forest
canopies have been widely reported, the patterns reported
here have not been examined in the literature.
5.5. Comparison of AMSR-E to the CMC snow product
Maximum October snow depth values over the Northern Hemisphere are 20–30 cm (Fig. 8). The AMSR-E
Author's personal copy
1020
A. Frei et al. / Advances in Space Research 50 (2012) 1007–1029
Fig. 7. Same as Fig. 5 except showing where MODIS disagrees with other two products.
product suggests more snow in Siberia than the CMC
product. AMSR-E overestimation with respect to CMC
over Siberia increases as the snow season progresses. In
January, snow depth differences between the two products
increase to 30–40 cm (Fig. 9). In April, the area over
which AMSR-E overestimates snow depth increases even
further with respect to January. In contrast, over other
regions AMSR-E tends to underestimate snow depth with
Author's personal copy
A. Frei et al. / Advances in Space Research 50 (2012) 1007–1029
1021
Fig. 8. Comparison of ASMR-E and CMC snow depth for October. (a) CMC estimated monthly mean snow depth. (b) AMSR-E estimated monthly
mean snow depth. (c) Difference (CMC – AMSR-E) in monthly mean snow depths.
respect to the CMC product, but these areas do not appear
to expand as the snow season progresses. These include the
Tibetan plateau and along the north-east coast of North
America (Fig. 10).
Histograms of the snow depth differences between the
two products are shown in Fig. 11. Overall, AMSR-E tends
to overestimate the values provided by CMC. While the
variance of the errors can be seen in the histogram plots,
perhaps a more informative number would be the coefficient of variation (Cv). Cv defined as the absolute value
of the standard deviation of the differences divided by the
mean CMC snow depth, provides an indication of how
large the differences are in comparison to the snow depth.
For example, a value of Cv = 1 means that the errors are
of the same magnitude as the mean depth; Cv = 0.1 means
that the errors or an order of magnitude less than the mean
snow depth. Cv values were calculated for each month
(Table 1). Cv values are highest in October, when depths
are small; lowest in January; and increase again in April.
As a snow pack ages, even under cold conditions without
additional precipitation, metamorphic processes lead to
grain size variations (such as depth hoar formation) that
tend to introduce errors in the passive microwave product.
Furthermore, as temperatures fluctuate and additional
Author's personal copy
1022
A. Frei et al. / Advances in Space Research 50 (2012) 1007–1029
Fig. 9. Same as Fig. 8 but for January.
precipitation events add fresh snow, snow packs can
develop a series of well defined layers of different grain sizes
that confound passive microwave based estimates of depth
and SWE. Ice layers, which can develop as a result of
melt-freeze and rain-freeze events, introduce additional
scattering and therefore additional uncertainty. Such complications, combined with the impact of vegetation,
especially vegetation that can change seasonally, can introduce a growing error in passive microwave retrievals as the
snow season progresses.
Improved confidence in our abilities to estimate snow
mass from satellites would support efforts to monitor the
fresh water flux into the Arctic Ocean. An order of magnitude estimate suggests that the volume of water in the snow
pack can play a significant role in the total annual river runoff into the Arctic Ocean of 4300–4800 km3 yr 1 (Arnell,
2005; Miller and Russell, 2000). Our AMSR-E based
(highly uncertain) estimate of the mean snow mass over
land surfaces during March (the month of maximum snow
mass) north of 60 N is 1600 km3. Frei et al. (2005) based
on the analysis of Brown et al. (2003) estimated the
observed mean snow volume over North America during
March to be 1500 km3, which was equal to the median
value estimated by a group of 18 climate models. This compares to a recent passive microwave-based estimate of
1400 km3 and 2300 km3 for mean North American
and Eurasian snow volumes, respectively (Biancamaria
et al., 2011). The errors associated with most of these
Author's personal copy
A. Frei et al. / Advances in Space Research 50 (2012) 1007–1029
1023
Fig. 10. Same as Fig. 8 but for April.
estimates are currently unknown, but they indicate that the
snowpack provides a significant fraction of the total river
runoff to the Arctic.
6. Discussion and conclusions
For most of the snow season and most regions there is
large-scale agreement amongst the products in identifying
the location of snow covered surfaces (i.e. snow extent,
regardless of snow depth) during clear sky conditions.
One exception to this is over central Asia. It is known that
passive microwave products identify snow on the Tibetan
Plateau and surrounding mountains when visible and
infrared products do not (Armstrong and Brodzik, 2001;
Basist et al., 1996). Because passive microwave retrieval
algorithms are calibrated at lower elevations, at these high
elevations the reduced atmosphere between the surface and
radiometer can result in retrieval errors (Savoie et al., 2009;
Wang and Tedesco, 2007). The second exception occurs
where snow is ephemeral, patchy, or wet. In such regions
the attenuation of the passive microwave signal, upon
which snow detection is based, is compensated for by emission from the surface or from liquid water in the snow pack
(Matzler, 1994; Ulaby and Stiles, 1980). Despite these difficulties, all estimates (discussed in the preceding section)
indicate that the snowpack is the source of a significant
portion of runoff into the Arctic basin.
The disagreements in snow extent during April are
greater than during October or January in terms of the percentage of available days during which one product differs
Author's personal copy
1024
A. Frei et al. / Advances in Space Research 50 (2012) 1007–1029
Table 1
Mean snow depth from CMC product; standard deviation of the
differences between the CMC and AMSR-E snow depths; and the
coefficient of variation. All values are averages of grid points across all
Northern Hemisphere land areas north of 30 N excluding the Greenland
ice sheet.
October
January
April
Fig. 11. Histograms of differences in snow depth (CMC – AMSR-E) for
all grid points for (a) October, (b) January, and (c) April.
from the other two. This is in agreement with Brown et al.
(2010), Frei and Lee (2010), and Brown et al. (2007), who
find differences between sensors during spring over North
American regions experiencing ablation, and indicates that
the wet snow during ablation is perhaps more of a hindrance
to the identification of snow from satellites than some of the
other confounding factors. However, during fall and winter
the evaluation is hampered by data availability problems
associated with cloudiness and solar illumination issues.
Our analysis also demonstrates that snow depths estimated by the Canadian Meteorological Centre product
l (mean CMC
snow depth)
(cm)
r (standard
deviation of
difference) (cm)
Cv abs(rl) (coefficient
of variation) (unitless)
5.0
22.5
22.9
4.15
7.18
12.33
0.83
0.32
0.54
and by the AMSR-E algorithm can differ substantially.
Although there are no absolute surface reference observations in most regions to determine which (if either) product
is correct, we know from experience as well as theory that
the passive microwave depth and SWE algorithms are inaccurate under certain conditions (Tedesco and Narvekar,
2010). Sources of error include: surface heterogeneity
within a passive microwave footprint; temporal and spatial
variability in grain size and snow density; obscuration of
snow by forests; masking of the passive microwave signal
by liquid water in the snow pack; and effects of atmospheric attenuation. The persistent underestimation by
AMSR-E with respect to CMC over some regions can be
partially explained by considering that snow depth over
many of those areas is above the ‘saturation’ depth to
which the passive microwave algorithm is sensitive (Derksen, 2008; Markus et al., 2006; Matzler, 1994; Schanda
et al., 1983); the presence of a high fraction of lakes over
the north east of North America is also believed to be a
source of error (Derksen et al., 2005a; Rees et al., 2006).
Another example is the overestimation of snow depth by
AMSR-E over northern Siberia, which can be attributed to
the limitation of the current AMSR-E algorithm to
account for the large grains that typically develop in snow
packs in this (and some other) regions (Clifford, 2010).
Over regions that develop and maintain a snow pack early
in the season, the snow tends to insulate the ground keeping it warm even as air temperatures fall, resulting in a
strong vertical temperature gradient in the snow pack. This
temperature gradient causes vertical energy and vapor
fluxes within the snow pack, the net effect of which is a
layer of depth hoar at the bottom of the snow pack (Jordan
et al., 2008). The large crystal sizes of depth hoar (5 mm)
cause increased scattering of microwave radiation resulting
in an overestimation of the snow pack by the passive
microwave algorithms.
Opportunities remain for the development of improved
snow products. For example, improvements can be made
with regard to the retrieval of snow amount from passive
microwave sensors (Tedesco et al., 2004) under forested
terrain (Derksen, 2008), the refinement of snow extent estimates from visible and infrared sensors (Parajka et al.,
2008), and the estimation of sub-grid scale information.
Tedesco and Miller (2007b) explore the relative merits of
Author's personal copy
A. Frei et al. / Advances in Space Research 50 (2012) 1007–1029
combining active and passive microwave retrievals, using a
MODIS snow product as their reference “truth.” A number of researchers are investigating the potential for finer
scale information on snow extent, amount, fractional snow
cover (Derksen et al., 2005b; Salomonson and Appel, 2004,
2006), snow melt (Wang et al., 2008), as well as on snow
pack properties (Kinar and Pomeroy, 2007; Nolin and
Dozier, 2000; Painter and Dozier, 2004; Painter et al.,
2003; Rango et al., 2000; Schmugge et al., 2002). Improvements in remotely sensed products that do not rely on the
assimilation of data or model results will come as a consequence of improved understanding of the interaction
between electromagnetic and geophysical parameters at
large spatial scales. In this context, a new operational algorithm based on the inversion of an electromagnetic model,
artificial neural networks and snow climatology currently
under evaluation may be capable of accounting for some
of these limitations.
One currently active area of research is the development
of combined products, which include in situ observations
and/or modeling results as well as remotely sensed information. One can identify advantages and disadvantages
to both combined and stand-alone remotely sensed products. While stand-alone remotely sensed products contain
inherent drawbacks as discussed here, at any time, either
in situ or remotely sensed data streams can fail, rendering
combined as well as stand-alone products vulnerable to
missing information. This is most critical for real- or
near-real time operational products, on which weather
forecast models or time-sensitive decisions rely.
Remote sensing of snow continues to contribute to our
understanding of Earth system processes. MODIS snow
products are valuable because they can provide high resolution snow estimates under cloud-free conditions using a
quantifiable algorithm. However, for climatological as well
as operational purposes, humans can integrate and filter
data from multiple sources and satellite images in ways
that fully automated methods are (at least currently)
unable to, and provide information for the entire land surface of the globe, regardless of the presence of clouds.
Thus, continuation of IMS, with its long record of snow
extent, is a priority. Considering the difficulties in determining SWE on a global scale from stand-alone remote
sensing products, it seems likely that combining multiple
sensors with station observations and/or models, such as
in the GlobSnow product will provide the best estimates
of SWE.
Acknowledgements
Frei is supported by the NASA Cryospheric Sciences Program award #NNX08AQ70G, and began work on this article while on sabbatical leave at the Climate Research
Division of Environment Canada in Downsview, Ontario.
M. Tedesco is supported by NASA Grant # NNX08AI02G.
D. Robinson acknowledges funding support from NASA
MEaSUREs award NNX08AP34A and NOAA Climate
1025
Program Office awards EA133E10SE2623 and NA08AR
4310678. Two anonymous reviewers made valuable contributions to, and helped clarify, our manuscript; and we thank
T. Estilow and Jeff Miller for contributions to the figures.
References
ACIA Impacts of a Warming Arctic: Arctic Climate Impact Assessment,
ACIA. Cambridge University Press, 140 p. Available from: <http://
www.amap.no/; http://amap.no/acia/>., 2004.
AMAP. SWIPA (Snow, Water, Ice, and Permafrost in the Arctic)
Executive Summary. Oslo, Norway, Arctic Monitoring and Assessment Program: 15 p., 2011.
Armstrong, R.L., Brodzik, M.J. Recent Northern Hemisphere snow
extent: a comparison of data derived from visible and microwave
satellite sensors. Geophysical Research Letters 28 (19), 3673–3676,
2001.
Arnell, N.W. Implications of climate change for freshwater inflows to the
Arctic Ocean. Journal of Geophysical Research 110, D07105, 2005.
Barnett, T.P., Adam, J.C., Lettenmaier, D.P. Potential impacts of a
warming climate on water availability in snow-dominated regions.
Nature 438, 303–309, 2005.
Barry, R. The role of snow and ice in the global climate system: a review.
Polar Geography 26 (3), 235–246, 2002.
Barry, R.G., Fallot, J.-M., Armstrong, R.L. Twentieth-century variability
in snow-cover conditions and approaches to detecting and monitoring
changes: status and prospects. Progress in Physical Geography 19 (4),
520–532, 1995.
Barry, R.G., Armstrong, R., Callaghan, T., Cherry, J., Gearhead, S.,
Nolin, A., Russell, D., Zockler, C. Snow. Global Outlook for Ice and
Snow. UNEP. Nairobi, Kenya, United Nations Environment Program: pp. 39–62, 2007.
Basist, A., Garrett, D., Ferraro, R., Grody, N., Mitchell, K. A comparison
between snow cover products derived from visible and microwave
satellite observations. Journal of Applied Meteorology 35, 163–177,
1996.
Betts, R.A. Offset of the potential carbon sink from boreal forestation by
decreases in surface albedo. Nature 408 (6809), 187–190, 2000.
Biancamaria, S., Cazevave, A., Mognard, N.M., Llovel, W., Frappart, F.
Satellite-based high latitude snow volume trend, variability and
contribution to sea level over 1989/2006. Global and Planetary
Change 75, 99–107, 2011.
Bitner, D., Carroll, T., Cline, D., Romanov, P. An assessment of the
differences between three satellite snow cover mapping techniques.
Hydrological Processes 16, 3723–3733, 2002.
Brasnett, B. A global analysis of snow depth for numerical weather
prediction. Journal of Applied Meteorology 38 (6), 726, 1999.
Brodzik, M.J., Knowles, K. EASE-Grid: A Versatile Set of Equal-area
Projections and Grids, in: M. Goodchild (Ed.), Discrete Global Grids.
National Center for Geographic Information Analysis, Santa Barbara,
California, USA, 2002.
Brodzik, M.J., Armstrong, R.A., Savoie, M. Global EASE-Grid 8-day
Blended SSM/I and MODIS Snow Cover, 2007. Available from:
<http://nsidc.org/data/docs/daac/nsidc0321_8day_ssmi_modis_blend/
index.html>.
Brown, R., Armstrong, R.L. Snow-cover data: measurement, products,
sources, in: Armstrong, R.L., Brun, E. (Eds.), Snow and Climate:
Physical Processes, Surface Energy Exchange and Modeling. Cambridge University Press, pp. 181–216, 2008.
Brown, R., Derksen, C., Wang, L. Assessment of spring snow cover
duration variability over northern Canada from satellite datasets.
Remote Sensing of Environment 111, 367–381, 2007.
Brown, R., Derksen, C., Wang, L. A multi-data set analysis of variability
and change in Arctic spring snow cover extent, 1967–2008. Geophysical Research Letters 115 (D16111), 16, 2010.
Brown, R.D. Northern Hemisphere snow cover variability and change,
1915–1997. Journal of Climate 13 (13), 2339–2355, 2000.
Author's personal copy
1026
A. Frei et al. / Advances in Space Research 50 (2012) 1007–1029
Brown, R.D., Braaten, R.O. Spatial and temporal variability of Canadian
monthly snow depths, 1946–1995. Atmosphere-Ocean 36, 37–45, 1998.
Brown, R.D., Mote, P.W. The response of Northern Hemisphere snow
cover to a changing climate. Journal of Climate 22, 2124–2145, 2009.
Brown, R.D., Brasnett, B., Robinson, D.A. Gridded North American
monthly snow depth and snow water equivalent for GCM evaluation.
Atmosphere-Ocean 41 (1), 1–14, 2003.
Brubaker, K.L., Pinker, R.T., Deviatova, E. Evaluation and comparison
of MODIS and IMS snow-cover estimates for the continental United
States using station data. Journal of Hydrometeorology 6, 1002–1017,
2005.
Carroll, T. Operational airborne measurements of snow water equivalent
and soil moisture using terrestrial gamma radiation in the United
States, in: Goodison, B., Barry, R.G., Dozier, J. (Eds.), Large Scale
Effects of Seasonal Snow Cover (Proceedings of the Vancouver
Symposium, August 1987). Vancouver, BC, IAHS, pp. 213–223, 1987.
Carroll, T., Cline, D., Fall, G., Nilsson, A., Li, L., Rost, A. NOHRSC
operations and the simulation of snow cover properties for the
coterminous US, in: 69th Annual Meeting of the Western Snow
Conference, Sun Valley, Idaho, pp. 2–14, 2001.
Chang, A.T.C., Foster, D.R., Hall, D.K. Nimbus-7 SMMR derived global
snow cover parameters. Annals of Glaciology 9, 39–44, 1987.
Chang, A.T.C., Foster, D.R., Hall, D.K. Effects of forest on the snow
parameters derived from microwave measurements during the
BOREAS winter field campaign. Hydrological Processes 10, 1565–
1574, 1996.
Chang, A.T.C., Gloersen, P., Schmugge, T.J., Wilheit, T., Zwally, H.J.
Microwave emission from snow and glacier ice. Journal of Glaciology
16 (74), 23–39, 1976.
Chang, A.T.C., Kelly, R.E.J., Josberger, E.G., Armstrong, R.L., Foster,
J.L., Mognard, N.M. Analysis of ground-measured and passivemicrowave-derived snow depth variations in midwinter across the
northern Great Plains. Journal of Hydrometeorology 6, 20–33, 2005.
Clifford, D. Global estimates of snow water equivalent from passive
microwave instruments: history, challenges and future developments.
International Journal of Remote Sensing 31 (14), 3707–3726, 2010.
Cohen, J. Snow cover and climate. Weather 49 (5), 150–156, 1994.
Derksen, C. The contribution of AMSR-E 18.7 and 10.7 GHz measurements to improved boreal forest snow water equivalent retrievals.
Remote Sensing of Environment 112, 2701–2710, 2008.
Derksen, C., Walker, A.E. Identification of systematic bias in the crossplatform (SMMR and SMM/I) EASE-Grid brightness temperature
time series. IEEE Transactions on Geoscience and Remote Sensing 41
(4), 910–915, 2003.
Derksen, C., Walker, A.E., Goodison, B.E. A comparison of 18 winter
seasons of in situ and passive microwave-derived snow water equivalent estimates in Western Canada. Remote Sensing of Environment
88, 271–282, 2003.
Derksen, C., Brown, R., Walker, A.E. Merging conventional (1915–92)
and passive microwave (1978–2002) estimates of snow extent and
water equivalent over central North America. Journal of Hydrometeorology 5, 850–861, 2004.
Derksen, C., Walker, A., Goodison, B. Evaluation of passive microwave
snow water equivalent retrievals across the boreal forest/tundra
transition of western Canada. Remote Sensing of Environment 96,
315–327, 2005a.
Derksen, C., Walker, A.E., Goodison, B.E., Strapp, J.W. Integrating
in situ and multiscale passive microwave data for estimation of subgrid
scale snow water equivalent distribution and variability. IEEE
Transactions on Geoscience and Remote Sensing 43 (5), 960–972,
2005b.
Derksen, C., Toose, P., Rees, A., Wang, L., English, M., Walker, A.,
Sturm, M. Development of a tundra-specific snow water equivalent
retrieval algorithm for satellite passive microwave data. Remote
Sensing of Environment 114, 1699–1709, 2010.
Dozier, J. Spectral signature of alpine snow cover from the landsat
thematic mapper. Remote Sensing of Environment 28, 9–22, 1989.
Drusch, M., Vasiljevic, D., Viterbo, P. ECMWF’s global snow analysis:
assessment and revision based on satellite observations. Journal of
Applied Meteorology 43, 1282–1294, 2004.
Dyer, J.L. Snow depth and streamflow relationships in large north
american watersheds. Journal of Geophysical Research 113, D18113,
2008.
Dyer, J.L., Mote, T.L. Spatial variability and trends in snow depth over
North America. Geophysical Research Letters 33, L16503, 2006.
Ellis, A.W., Leathers, D.J. Analysis of cold airmass temperature modification across the US great plains as a consequence of snow depth and
albedo. Journal of Applied Meteorology 38 (6), 696–711, 1999.
Foster, J., Liston, G., Koster, R., Essery, R., Behr, H., Dumenil, L.,
Verseghy, D., Thompson, S., Pollard, D., Cohen, J. Snow cover and
snow mass intercomparison of general circulation models and remotely
sensed datasets. Journal of Climate 9, 409–426, 1996.
Foster, J.L., Chang, A.T.C., Hall, D.K. Comparison of snow mass
estimates from a prototype passive microwave snow algorithm, a
revised algorithm and a snow depth climatology. Remote Sensing of
Environment 62, 132–142, 1997.
Foster, J.L., Hall, D.K., Kelly, R.E.J., Chiu, L. Seasonal snow extent and
snow mass in South America using SMMR and SSM/I passive
microwave data (1979-2006). Remote Sensing of Environment 113,
291–305, 2008.
Foster, J.L., Hall, D.K., Eylander, J.B., Riggs, G.A., Nghiem, S.V.,
Tedesco, M., Kim, E., Montesano, P.M., Kelly, R.E.J., Casey, K.A.,
Choudhury, B. A blended global snow product using visible, passive
microwave and scatterometer data. International Journal of Remote
Sensing 32 (5–6), 1371–1395, 2011.
Francis, J.A., White, D., Cassano, J., Gutkowski, W.J., Hinzman, L.D.,
Holland, M., Steele, M., Vorosmarty, C.J. An Arctic hydrologic
system in transition: feedbacks and impacts on terrestrial, marine, and
human life. Journal of Geophysical Research 115, G04019, 2009.
Frei, A., Gong, G. Decadal to century scale trends in North American
snow extent in coupled atmosphere-ocean general circulation models.
Geophysical Research Letters 32 (18), L18502, 5 pp., 2005.
Frei, A., Lee, S. A comparison of optical-band based snow extent
products during spring over North America. Remote Sensing of
Environment 114, 1940–1948, 2010.
Frei, A., Robinson, D.A., Hughes, M.G. North American snow extent:
1900–1994. International Journal of Climatology 19, 1517–1534, 1999.
Frei, A., Miller, J.A., Robinson, D.A. Improved simulations of snow
extent in the second phase of the Atmospheric Model Intercomparison
Project (AMIP-2). Journal of Geophysical Research – Atmospheres
108 (D12), 4369, doi:10.1029/2002JD003030, 2003.
Frei, A., Brown, R., Miller, J.A., Robinson, D.A. Snow mass over North
America: observations and results from the second phase of the
Atmospheric Model Intercomparison Project (AMIP-2). Journal of
Hydrometeorology 6 (5), 681–695, 2005.
Friedl, M.A., McIver, D.K., Hodges, J.C.F., Zhang, X.Y., Muchoney, D.,
Strahler, A.H., Woodcock, C.E., Gopal, S., Schneider, A., Cooper, A.,
Baccini, A., Gao, F., Schaaf, C. Global land cover mapping from
MODIS: algorithms and early results. Remote Sensing of Environment
83 (1–2), 287–302, 2002.
Ge, Y., Gong, G. Observed Inconsistencies between snow extent and snow
depth variability at regional/continental scales. Journal of Climate 21,
1066–1082, 2008.
Goita, K., Walker, A., Goodison, B. Algorithm development for the
estimation of snow water equivalent in the boreal forest using passive
microwave data. International Journal of Remote Sensing 24, 1097–
1102, 2003.
Goodison, B., Ferguson, H.L., McKay, G.A. Measurement and data
analysis, in: Gray, D.M., Male, D.M. (Eds.), Handbook of Snow:
Principles, Processes, Management & Use. Pergamon Press Canada
Ltd., Toronto, pp. 91–274, 1981.
Goodison, B.E., Walker, A.E. Canadian development and use of snow
cover information from passive microwave satellite data, in: Choudhury, B., Kerr, Y., Njoku, E., Pampaloni, P. (Eds.), Passive Microwave
Author's personal copy
A. Frei et al. / Advances in Space Research 50 (2012) 1007–1029
Remote Sensing of Land–Atmosphere Interactions, Utrecht, Netherlands, VSP BV, pp. 245–262, 1995.
Graybeal, D., Leathers, D. Snowmelt-related flood risk in Appalachia:
first estimates from a historical snow climatology. Journal of Applied
Meteorology and Climatology 45, 178–193, 2006.
Grody, N. Relationship between snow parameters and microwave
satellite measurements: theory compared with AMSU observations
from 23 to 150 GHz. Journal of Geophysical Research 113, D22108,
2008.
Grody, N.C., Basist, A.N. Global identification of snowcover using SSM/I
measurements. IEEE Transactions on Geoscience and Remote Sensing
34 (1), 237–249, 1996.
Groisman, P.Y., Karl, T.R., Knight, R.W. Observed impact of snow cover
on the heat balance and the rise of continental spring temperatures.
Science 263, 199–200, 1994.
Grundstein, A., Mote, T., Leathers, D.J. A hybrid climatology of snow
water equivalent over the northern Great Plains of the United States.
Polar Geography 26 (3), 187–209, 2002.
Gutzler, D.S., Rosen, R.D. Interannual variability of wintertime snow
cover across the Northern Hemisphere. Journal of Climate 5, 1441–
1447, 1992.
Hall, D.K., Riggs, G.A. Accuracy assessment of the MODIS snow
products. Hydrological Processes 21, 1534–1547, 2007.
Hall, D.K., Riggs, G.A., Salomonson, V.V. Development of methods for
mapping global snow cover using moderate resolution imaging
spectroradiometer data. Remote Sensing of Environment 54, 127–
140, 1995.
Hall, D.K., Kelly, R.E.J., Foster, J.L., Chang, A.T.C. Estimation of Snow
Extent and Snow Properties, in: Anderson, M.G. (Ed.), Encyclopedia
of Hydrological Sciences. Wiley, Berlin, pp. 811–829, 2005.
Hall, D.K., Riggs, G.A., Foster, J.L., Kumar, S.V. Development and
evaluation of a cloud-gap-filled MODIS daily snow-cover product.
Remote Sensing of Environment 114 (3), 496–503, 2010.
Hall, D.K., Riggs, G.A., Salomonson, V.V., DiGirolamo, N.E., Bayr, K.J.
MODIS snow-cover products. Remote Sensing of Environment 83,
181–194, 2002.
Helfrich, S.R., McNamara, D., Ramsay, B.H., Baldwin, T., Kasheta, T.
Enhancements to and forthcoming developments in the Interactive
Multisensor Snow and Ice Mapping System (IMS). Hydrological
Processes 21, 1576–1586, 2007.
Hinzman, L.D., Bettez, N.D., Bolton, W.R., Chapin, F.S., Dyurgerov,
M.B., Fastie, C., Griffith, B., Hollister, R.D., Hope, A., Huntington,
H.P., Jensen, A.M., Jia, G.J., Jorgenson, T., Kane, D., Klein, D.R.,
Kofinas, G., Lynch, A.H., Lloyd, A.H., xMcGuire, A.H., Nelson,
F.E., Oechel, W.C., Osterkamp, T.E., Racine, C., Romanovsky, V.E.,
Stone, R.S., Stow, D.A., Sturm, M., Tweedie, C., Vourlitis, G.L.,
Walker, M.D., Walker, D.A., Webber, P.J., Welker, J.M., Winker,
K.S., Yoshikawa, K. Evidence and implications of recent climate
change in northern Alaska and other arctic regions. Climatic Change
72, 251–298, 2005.
IPCC. Climate Change 2007: The Physical Science Basis. Contribution of
Working Group I to the Fourth Assessment Report of the Intergovernmental Panel on Climate Change, in: S.D. Soloman et al.,
Intergovernmental Panel on Climate Change. Cambridge University
Press.
Jones, H.G., Pomeroy, J.W., Walker, D.A., Hoham, R.W. Snow Ecology:
An Interdisciplinary Examination of Snow-Covered Ecosystems.
Cambridge University Press, UK, Cambridge, 2001.
Jordan, R. A one-dimensional temperature model for a snow cover:
technical documentation for SNTHERM.89. US Army Corps of
Engineers, Cold Regions Research and Engineering Laboratory,
Hanover, NH, 1991.
Jordan, R.E., Albert, m.R., Brun, E. Physical properties within the snow
cover and their parameterization, in: Armstrong, R.A., Brun, E.
(Eds.), Snow and Climate: Physical Processes, Surface Energy
Exchange and Modeling. Cambridge University Press, Cambridge,
UK, p. p. 22, 2008.
1027
Kelly, R.E.J. The AMSR-E snow depth algorithm: description and initial
results. Journal of the Remote Sensing Society of Japan 29 (1), 307–
317, 2009.
Kelly, R.E.J., Chang, A.T.C., Tsang, L., Foster, J.L. Development of a
prototype AMSR-E global snow area and depth algorithm. IEEE
Transactions on Geoscience and Remote Sensing 41 (2), 230–242,
2003.
Kelly, R.E.J., Chang, A.T.C., Foster, J.L., Hall, D.K. Using remote
sensing and spatial models to monitor snow depth and snow water
equivalent, in: Kelly, R.E.J., Drake, N.A., Barr, S.L. (Eds.), Spatial
Modeling of the Terrestrial Environment. John Wiley and Sons, Ltd.,
Chichester, England, pp. 5–57, 2004.
Kinar, N.J., Pomeroy, J.W. Determining snow water equivalent by
acoustic sounding. Hydrological Processes 21 (19), 2623–2640, 2007.
Klein, A.G., Stroeve, J. Development and validation of a snow albedo
algorithm for the MODIS instrument. Annals of Glaciology 34, 45–52,
2002.
Klein, A.G., Hall, D.K., Riggs, G.A. Improving snow cover mapping in
forests through the use of a canopy reflectance model. Hydrological
Processes 12, 1723–1744, 1998.
Langham, E.J. Physics and properties of snow cover, in: Gray, D.M.,
Male, D.M. (Eds.), Handbook of Snow: Principles, Processes, Management & Use. Pergamon Press Canada Ltd., Toronto, pp. 275–337,
1981.
Lawrence, D.M., Slater, A.G. A projection of severe near-surface
permafrost degradation during the 21st century. Geophysical Research
Letters 32 (L24401), doi:10.1029/2005GL025080, 2005.
Leathers, D.J., Robinson, D.A. The association between extremes in
North American snow cover extent and United States temperature.
Journal of Climate 6, 1345–1355, 1993.
Leathers, D.J., Kluck, D.R., Kroczynski, S. The severe flooding event of
January 1996 across north central Pennsylvania. Bulletin of the
American Meteorological Society 79 (5), 785–797, 1998.
Lenton, T.M., Held, H., Kriegler, E., Hall, J.W., Lucht, W., Rahmstorf,
S., Schellnhuber, H.J. Tipping elements in the Earth’s climate system.
Proceedings of the National Academy of Sciences 105 (6), 1786–1796,
2008.
MacKay, M.D., Bartlett, P.A., Chan, E., Derksen, C., Guo, S., Leighton,
H. On the simulation of regional scale sublimation over boreal and
agricultural landscapes in a climate model. Atmosphere-Ocean 44 (3),
289–304, 2006.
Markus, T., Powell, D.C., Wang, J.R. Sensitivity of passive microwave
snow depth retrievals to weather effects and snow evolution. IEEE
Transactions on Geoscience and Remote Sensing 44 (1), 68–77, 2006.
Matson, M., Wiesnet, D.R. New data base for climate studies. Nature 289,
451–456, 1981.
Matzler, C. Passive microwave signatures of landscapes in winter.
Meteorological and Atmospheric Physics 54, 241–260, 1994.
Metsamaki, S.J., Anttila, S.T., Markus, H.J., Vepsalainen, J.M. A feasible
method for fractional snow cover mapping in boreal zone based on a
reflectance model. Remote Sensing of Environment 95, 77–95, 2005.
Mialon, A., Fily, M., Royer, A. Seasonal snow cover extent from
microwave remote sensing data: comparison with existing ground and
satellite based measurements, in: European Association of Remote
Sensing Laboratories (EARSeL), Strasbourg, France, pp. 215–225,
2005.
Miller, J.R., Russell, G.L. Projected impact of climate change on the
freshwater and salt budgets of the Arctic Ocean by a global climate
model. Geophysical Research Letters 27 (8), 1183–1186, 2000.
Mote, P.W. Climate-driven variability and trends in mountain snowpack
in western North America. Journal of Climate 19, 6209–6220, 2006.
Mote, P.W., Hamlet, A.F., Clark, M.P., Lettenmaier, D.P. Declining
mountain snowpack in western North America. Bulletin of the
American Meteorological Society 86 (1), 39–49, 2005.
Mote, T.L. On the role of snow cover in depressing air temperature.
Journal of Applied Meteorology and Climatology 47, 2008–2022,
2008.
Author's personal copy
1028
A. Frei et al. / Advances in Space Research 50 (2012) 1007–1029
Mote, T.L., Grundstein, A.J., Leathers, D.J., Robinson, D.A. A
comparison of modeled, remotely sensed, and measured snow water
equivalent in the northern Great Plains. Journal of Applied Meteorology 43, 1887–1898, 2003.
Nghiem, S.V., Steffen, K., Kwok, R., Tsai, W.Y. Detection of snowmelt
regions on the Greenland ice sheet using diurnal backscatter change.
Journal of Glaciology 47 (159), 539–547, 2001.
Nolin, A.W. Towards retrieval of forest cover density over snow from the
Multi-angle Imaging SpectroRadiometer (MISR). Hydrological Processes 18, 3623–3636, 2004.
Nolin, A.W. Recent advances in remote sensing of seasonal snow. Journal
of Glaciology 56 (200), 1141–1150, 2010.
Nolin, A.W., Dozier, J. A hyperspectral method for remotely sensing the
grain size of snow. Remote Sensing of Environment 74 (2), 207–216,
2000.
Nolin, A.W., Daly, C. Mapping “at-risk” snow in the Pacific Northwest,
USA. Journal of Hydrometeorology 7, 1166–1173, 2006.
Painter, T.H., Dozier, J. The effect of anisotropic reflectance on imaging
spectroscopy of snow properties. Remote Sensing of Environment 89,
409–422, 2004.
Painter, T.H., Dozier, J., Roberts, D.A., Davis, R.E., Green, R.O.
Retrieval of subpixel snow-covered area and grain size from imaging
spectrometer. Remote Sensing of Environment 85, 64–77, 2003.
Painter, T.H., Rittger, K., McKenzie, C., Slaughter, P., Davis, R.E.,
Dozier, J. Retrieval of subpixel snow covered area, grain size, and
albedo from MODIS. Remote Sensing of Environment 113, 868–879,
2009.
Parajka, J., Bloschl, G. Spatio-temporal combination of MODIS images –
potential for snow cover mapping. Water Resources Research 44,
W03406, 2008.
Peterson, B., Holmes, J.R.M., McClelland, J.W., Vorosmarty, C.J.,
Lammers, R.B., Shiklomanov, A.I., Shiklomanov, I.A., Rahmstorf,
S. Increasing river discharge to the Arctic Ocean. Science 298 (5601),
2171–2173, 2002.
Pulliainen, J. Mapping of snow water equivalent and snow depth in boreal
and sub-Arctic zones by assimilating space-borne microwave radiometer data and ground-based observations. Remote Sensing of Environment 101 (2), 257–269, 2006.
Pulliainen, J. Overview on GlobSnow Project: key questions to be resolved
(coverage, aggregation, validation, towards a future ECV), in: The 1st
ESA DUE GlobSnow User Workshop, Innsbruck, Austria, 2010.
Pulliainen, J., Hallikainen, M. Retrieval of regional snow water equivalent
from space-borne passive microwave observations. Remote Sensing of
Environment 75 (1), 76–85, 2001.
Rahmstorf, S. The thermohaline ocean circulation: a system with
dangerous thresholds? Climatic Change 46 (3), 247–256, 2000.
Raisanen, J. Warmer climates: less or more snow? Climate Dynamics 30,
307–319, 2007.
Ramsay, B.H. The interactive multisensor snow and ice mapping system.
Hydrological Processes 12, 1537–1546, 1998.
Ramsay, B.H. Prospects for the Interactive Multisensor Snow and Ice
Mapping System (IMS), in: 57th Eastern Snow Conference, Syracuse,
New York, USA, 2000.
Rango, A., Walker, A.E., Goodison, B.E. Snow and ice, in: Schultz, G.A.,
Engman, E.T. (Eds.), Remote Sensing in Hydrology and Water
Management. Springer, Berlin, pp. 39–262, 2000.
Rawlins, M., Steele, M., Holland, M., Adam, J.C., Cherry, J., Francis,
J.A., Groisman, P.Y., Hinzman, L.D., Huntington, T.G., Kane, D.,
Kimball, J., Kwok, R., Lammers, R.B., Lee, C., Lettenmaier, D.P.,
McDonald, K., Podest, E., Pundsack, J., Rudels, B., Serreze, M.C.,
Shiklomanov, A.I., Skagseth, O., Troy, T., Vorosmarty, C.J., Wensnahan, M., Wood, E., Woodgate, R., Yang, D., Zhang, K., Zhang, T.
Analysis of the Arctic system for freshwater cycle intensification:
observations and expectations. Journal of Climate 23, 5715–5737,
2010.
Rees, A., Derksen, C., English, M., Walker, A., Duguay, C. Uncertainty
in snow mass retrievals from satellite passive microwave data in lake-
rich high-latitude environments. Hydrological Processes 20, 1019–
1022, 2006.
Rennermalm, A.K., Wood, E.F., Weaver, A.J., Eby, M., Dery, S.J.
Relative sensitivity of the Atlantic meridional overturning circulation
to river discharge into Hudson Bay and the Arctic Ocean. Journal of
Geophysical Research 112 (G04S48), 2007.
Riggs, G.A., Hall, D.K. Reduction of Cloud Obscuration in the MODIS
Snow Data Product, in: Eastern Snow Conference, Stowe, Vermont,
pp. 205–212, 2002.
Riggs, G.A., Hall, D.K. MODIS Snow cover Algorithms and Products –
Plans for Next Version, in: 68th Annual Eastern Snow Conference,
Montreal, QC, Canada, in press.
Riggs, G.A., Digirolamo, N., Hall, D.K. Comparison of MODIS daily
global fractional snow cover maps at 0.05- and 0.25-Degree resolutions, in: 62nd Eastern Snow Conference, Waterloo, ON, Canada, pp.
21–27, 2005.
Riggs, G.A., Hall, D.K., Salomonson, V.V. “MODIS Snow Products
Users’ Guide”, 2006. Available from: <http://modis-snow-ice.gsfc.nasa.gov/sugkc2.html>.
Robinson, D., Kukla, G. Maximum surface albedo of seasonally snow
covered lands in the Northern Hemisphere. Journal of Climate and
Applied Meteorology 24, 402–411, 1985.
Robinson, D.A., Frei, A. Seasonal variability of Northern Hemisphere
snow extent using visible satellite data. Professional Geographer 52 (2),
307–315, 2000.
Robinson, D.A., Estilow, T. A Northern Hemisphere Snow Extent
Climate Data Record, in: Fall Meeting of the American Geophysical
Union, San Francisco, CA, 2008.
Robinson, D.A., Dewey, K.F., Heim, R.R.J. Global snow cover monitoring: an update. Bulletin of the American Meteorological Society 74
(9), 1689–1696, 1993.
Robinson, D.A., Tarpley, J.D., Ramsay, B.H. Transition from NOAA
weekly to daily hemispheric snow charts, in: 10th Symposium on
Global Change Studies, Dallas, TX. American Meteorological Society,
1999.
Roesch, A., Gilgen, H., Ohmura, A. Assessment of GCM simulated snow
albedo using direct observations. Climate Dynamics 15 (6), 405, 1999.
Romanov, P., Gutman, G., Csiszar, I. Satellite-derived snow cover maps
for North America: accuracy assessment. Advances in Space Research
30 (11), 2455–2460, 2002.
Romanov, P., Tarpley, D., Gutman, G., Carroll, T. Mapping and
monitoring of the snow cover fraction over North America. Journal of
Geophysical Research 108 (D16), 2003.
Romanovsky, V.E., Smith, S.L., Christiansen, H.H. Permafrost thermal
state in the Polar Northern Hemisphere during the International Polar
Year 2007–2009: a synthesis. Permafrost and Periglacial Processes 21,
106–116, 2010.
Salomonson, V.V., Appel, I. Estimating fractional snow coverage from
MODIS using the Normalized Difference Snow Index (NDSI). Remote
Sensing of Environment 89, 351–360, 2004.
Salomonson, V.V., Appel, I. Development of the Aqua MODIS NDSI
fractional snow cover algorithm and validation studies. IEEE Transactions on Geoscience and Remote Sensing 44 (7), 1747–1756, 2006.
Savoie, M.H., Wang, J., Brodzik, M.J., Armstrong, R.L. Improved Snow
Cover Retrievals from Satellite Passive Microwave Data over the Tibet
Plateau: the Need for Atmospheric Corrections over High Elevations
(Poster), in: American Geophysical Union, Fall Meeting Supplement
Abstract C23A-0942, 2007.
Savoie, M.H., Armstrong, R.A., Brodzik, M.J., Wang, J.R. Atmospheric
corrections for improved satellite passive microwave snow cover
retrievals over the Tibet Plateau. Remote Sensing of Environment 113,
2661–2669, 2009.
Schanda, E., Matzler, C., Kunzi, K. Microwave remote sensing of snow
cover. International Journal of Remote Sensing 4, 149–158, 1983.
Scherer, D., Hall, D.K., Hochschild, V., Konig, M., Winther, J.-G.,
Duguay, C.R., Pivot, F., Matzler, C., Rau, F., Seidel, K., Solberg, R.,
Walker, A.E. Remote Sensing of Snow. Remote Sensing in Northern
Author's personal copy
A. Frei et al. / Advances in Space Research 50 (2012) 1007–1029
Hydrology: Measuring Environmental Change, C. R. a. P. Duguay,
Alain, Washington, DC. American Geophysical Union, p. 163, 2005.
Schmugge, T.J., Kustas, W.P., Ritchie, J.C., Jackson, T.J., Rango, A.
Remote sensing in hydrology. Advances in Water Resources 25, 1367–
1385, 2002.
Schuur, E.A.G., Bockheim, J., Canadell, J.G., Euskirchen, E., Field, C.B.,
Goryachkin, S.V., Hagemann, S., Kuhry, P., Lafleur, P.M., Lee, H.,
Mazhitova, G., Nelson, F.E., Rinke, A., Romanovsky, V.E., Shiklomanov, N., Tarnocai, C., Venevsky, S., Vogel, J.G., Zimov, S.A.
Vulnerability of permafrost carbon to climate change: implications for
the global carbon cycle. Bioscience 58 (8), 701–714, 2008.
Siljamo, N., Hyvarinen, O. New geostationary satellite-based snow-cover
algorithm. Journal of Applied Meteorology and Climatology 50, 1275–
1290, 2011.
Sturm, M., Holmgren, J., Liston, G.E. A seasonal snow cover classification system for local to global applications. Journal of Climate 8 (5),
1261–1283, 1995.
Tait, A. Estimation of snow water equivalent using passive microwave
radiation data. Remote Sensing of Environment 64, 286–291, 1998.
Tait, A., Armstrong, R. Evaluation of SMMR satellite-derived snow
depth using ground-based measurements. International Journal of
Remote Sensing 17 (4), 657–665, 1996.
Tait, A.B., Hall, D.K., Foster, J.L., Armstrong, R.L. Utilizing multiple
datasets for snow-cover mapping. Remote Sensing of Environment 72,
111–126, 2000.
Takala, M., Luojus, K., Pulliainen, J., Derksen, C., Lemmetyinen, J.,
Karna, J.-P., Koskinen, J., Bojkov, B. Estimating Northern Hemisphere snow water equivalent for climate research through assimilation
of space-borne radiometer data and ground-based measurements.
Remote Sensing of Environment 115, 3517–3529, 2011.
Tedesco, M., Kim, E.J. Inter-comparison of electromagnetic models for
passive microwave remote sensing of snow. IEEE TGARS Special
Issue ‘IGARSS 2005’ 44 (10), 2654–2666, 2006.
Tedesco, M., Miller, J. Observations and statistical analysis of combined
active-passive microwave space-borne data and snow depth at large
spatial scales. Remote Sensing of Environment 111 (2–3), 382–397,
2007a.
Tedesco, M., Miller, J. Northern Hemisphere snow-covered area mapping:
optical versus active and passive microwave data. IEEE Geoscience
and Remote Sensing Letters 4 (2), 221–225, 2007b.
Tedesco, M., Narvekar Assessment of the NASA AMSR-E SWE product.
IEEE Journal of Selected Topics in Applied Earth Observations and
Remote Sensing 3 (1), 141–159, 2010.
Tedesco, M., Kim, E.J., Gasiewski, A., Stankov, B. Analysis of multi-scale
radiometric data collected during the Cold Land Processes Experiment-1 (CLPX-1). Geophysical Research Letters 32 (L18501), 4, 2005.
1029
Tedesco, M., Kelly, R.E.J., Foster, J.L., Chang, A.T.C. AMSR-E/Aqua
Daily L3 Global Snow Water Equivalent EASE-Grids V002, National
Snow and Ice Data Center Digital Media, Boulder Colorado, USA,
2011.
Tedesco, M., Pullianinen, J., Takala, M., Hallikainen, M., Pampaloni, P.
Artificial neural network-based techniques for the retrieval of SWE
and snow depth from SSM/I data. Remote Sensing of Environment 90
(1), 76–85, 2004.
Todhunter, P.E. A hydroclimatological analysis of the Red River of the
North snowmelt flood catastrophe of 1997. Journal of the American
Water Resources Association 37 (5), 1263–1278, 2001.
Tong, J., Velicogna, I. A comparison of AMSR-E/Aqua snow products
with in situ observations and MODIS snow cover products in
the Mackenzie River Basin, Canada. Remote Sensing 2, 2313–2322,
2010.
Tong, J., Dery, S.J., Jackson, P.L., Derksen, C. Testing snow water
equivalent retrieval algorithms for passive microwave remote sensing
in an alpine watershed of western Canada. Canadian Journal of
Remote Sensing 36 (Supplement 1), S74–S86, 2010a.
Tong, J., Dery, S.J., Jackson, P.L., Derksen, C. Snow distribution from
SSM/I and its relationships to the hydroclimatology of the Mackenzie
River Basin, Canada. Advances in Water Resources 33 (6), 667–677,
2010b.
Ulaby, F.T., Stiles, W.H. The active and passive microwave response to
snow parameters 2. Water equivalent of dry snow. Journal of
Geophysical Research 85 (C2), 1045–1049, 1980.
Walsh, J.E. Snow cover and atmospheric variability. American Scientist
72, 50–57, 1984.
Wang, J.R., Tedesco, M. Identification of atmospheric influences on the
estimation of snow water equivalent from AMSR-E measurements.
Remote Sensing of Environment 111 (2–3), 398–408, 2007.
Wang, L., Derksen, C., Brown, R. Detection of pan-Arctic terrestrial
snowmelt from QuikSCAT, 2000–2005. Remote Sensing of Environment 112 (10), 3794–3805, 2008.
Ye, H., Mather, J.R. Polar snow cover changes and global warming.
International Journal of Climatology 17, 155–162, 1997.
Ye, H., Cho, H.-R., Gustafson, P.E. The changes in Russian winter snow
accumulation during 1936–83 and its spatial patterns. Journal of
Climate 11, 856–863, 1998.
Zhang, T. Influence of the seasonal snow cover on the ground
thermal regime: an overview. Reviews of Geophysics 43, RG2003,
2005.
Zhang, X., Zwiers, F.W., Hegerl, G.C., Lambert, F.H., Gillett, N.P.,
Soloman, S., Stott, P.A., Nozawa, T. Detection of human influence on
20th-century precipitation trends. Nature 448, 461–466, 2007.